Agreed! The removal of this feature is enough for me to have to rethink next years cursor roll out.. I need to be able to dependably dictate workflow modes and their capability for development resources, nevermind models and costs. I guess if they are only looking for short term vibe coder subscriptions, this is a good way to go.. lol but if they want to maintain long term contracts with development shops, this will be a deal breaker.. either immediately if the customer knows what they are doing, or eventually, when the code has become a mangled mess and usage costs are skyrocketing and executive leadership is fed up with the developers not having confidence in timelines, deliveries or costs. Why not just add agent_modes to .cursor and be done with it? let them be project set and move on? Give enterprises what they need, if they know how to use it, and leave the premium upsell services for short term vibe coders that need them. best of both worlds..
The core value of Custom Modes was controlling the LLM experience without relying on the LLM. For one example, manually disabling an MCP is not the same as telling an LLM to “please not use it” in a Command’s Markdown file.
Additionally, a mode and all its customizations could be set once and persist across repeated prompts and Agent conversations. A Command must be manually typed again every single time. This makes it much more likely for the human to forget to type it and waste time (and tokens) on an incorrect prompt.
A key differentiator of Cursor versus other AI-assisted coding tools has long been the rich deterministic tuning available that takes place before LLMs start processing tokens. Examples that come to mind include Rule activation criteria, the context pills… and Custom Modes.
Please bring them back; I cannot achieve with Commands what I was doing with Custom Modes.
Feels good when you have to make a feature request to return a feature… At this point, i am 100% sure that cursor have no clear direction on what they want and how do they want to reach it.
I am so angry about this! I am getting my entire team off Cursor and also unrecruiting the 100s I have convinced that Cursor is the way to go. This is stupid. My first move. Get the message out to my followers on LinkedIn.
———-
Custom modes are gone in Cursor, and I don’t even know what today. I have never been one to go crazy about these things, but I feel weirdly betrayed this morning.
If you were really into agentic coding, custom modes were the way you made writing production code a thing in YOLO mode. Infrastructure as Code, APIs, Electron apps, Extensions, full AI pipelines, a freaking breeze with custom modes.
The product manager is me cannot fathom why they would do that in the manner they did. The other side? The person who runs venture studios and startup accelerators get the reason (people who use this are not ICP and they are not maxing out revenue, because we ahhh know what we are doing and don’t waste tokens) why but neither side gets the how they executed this move.
I am officially looking for a new IDE for not only me, the entire squad and everyone that I have personally convinced to come over to the Cursor side.
This is just unbelievable. This feels like the ultimate rug pool.
I was also frustrated to see this removed. However I’ve found I can get pretty much the exact same result using custom slash commands. Just put all your mode instructions into a cursor command and “/” mention it at the start of your chat.
You can also add a cursor rule to use custom commands when conditions you define are met i.e. “always”, “when the user mentions x, y, z” etc.
I think in many ways its similar.. but there is a certain level of control inherent to programmatically forbidding the ai agent to call tools and therefore save tokens used in both unwanted tool calls as well as the processing of the results of those unwanted tool calls..
This simple feature saved mindful users significant volumes of tokens. It also provided piece of mind to the end user that in that mode, the agent was constrained. This in turn supported the creation of workflows to optimize the process by those who were willing to leverage it to make repeatable processes for their teams…
Workflows, that I would argue would only speed AI augmented development and community adoption in the future.. Applying constraints within a rule works most the time.. but its not 100%, custom modes gave us that.
Today is December the second, yet no replies or updates about it, will Team Cursor ever respond or anything about Custom Modes anymore? At least as gift for the holiday
I’m so sure this will not harm the product nor decrease any value or peformance
Condor responded on the bug ticket “they are working on something… but its not ready”. Which might be … something . But I guess what really burns me here is product management 101, you don’t burn features without a plan… kind of like you don’t develop a feature without a plan… Cursor, your in the big leagues now… your going to have to start acting like it.
This was the intention - migrating custom modes to slash commands, which have been much more widely adopted and reliable (custom modes allowed picking tools, but many custom modes never configured tools, which is why we went down the custom modes to commands migration path; also, removing tools from models sometimes lead them to worse results).
Slash commands don’t last for the entire chat. That is the biggest problem. Having to do that every time is a waste of time. Additionally, having the mode shown in the UI helps me differentiate between which agent I am using (I usually have multiple chats open at once with different modes). In addition, slash commands don’t automatically switch the model to the one you want for that system prompt. You’ve traded tools designed for organized code generation for tools designed for yolo vibe coding. Ridiculous
I think the problem is Cursor has grown too quick, and they dont have product team or the product culture established.. they have product developers. It has gotten them this far.. But it won’t continue to work if they don’t have someone at the reigns to prioritize features and capabilities, as well as establish their brand now that they have established their differentiator. Instead, they seam to be “feature hunting” without having clear goals to support their differing userbase. Enterprise development shops, small team indies, 1 man freelancers, hobbyists and vibe coders with close to no code experience, they all have different needs and wants from the product. Right now, cursor appears to be addressing all of them badly instead of zeroing in and stabilizing on what they have been doing right, instead of chasing the new hot feature. Its a bummer, because I really think they have something here that could be truly industry breaking… if they don’t seriously hurt themselves (or their customers) through self-inflicted cuts.
I agree with Jordan. Slash commands do not last. Custom mode persistence with visual feedback and separation allowed me to create a full workflow for team rollout, 2 weeks before I do the rollout this happens. I am very frustrated.
Sad part is, I asked cursor plan mode chat, what the hell to do now, it pretty much said advocate to cursor to get this back as there is no alternative to easily get my custom mode functionality back.
I added a new topic here, additionally I sent an email to support.